hw/ppc/spapr: Make sure to close the htab_fd when migration is canceled
authorThomas Huth <thuth@redhat.com>
Thu, 21 Jul 2016 09:21:34 +0000 (11:21 +0200)
committerDavid Gibson <david@gibson.dropbear.id.au>
Mon, 25 Jul 2016 00:19:30 +0000 (10:19 +1000)
When canceling a migration process, we currently do not close the
HTAB migration file descriptor since htab_save_complete() is never
called in that case. So we leave the migration process with a
dangling htab_fd value around, and this causes any further migration
attempts to fail. To fix this issue, simply make sure that the
htab_fd is closed during the migration cleanup stage. And since the
cleanup() function is also called when migration succeeds, we can
also remove the call to close_htab_fd() from the htab_save_complete()
function.

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1354341
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
hw/ppc/spapr.c

index 7f33a1b2b57d8331f184d63e9fa97afcee7a2a0f..9193ac2c122bc3c4f19bd61266c2321e08310b54 100644 (file)
@@ -1512,7 +1512,6 @@ static int htab_save_complete(QEMUFile *f, void *opaque)
         if (rc < 0) {
             return rc;
         }
-        close_htab_fd(spapr);
     } else {
         if (spapr->htab_first_pass) {
             htab_save_first_pass(f, spapr, -1);
@@ -1614,10 +1613,18 @@ static int htab_load(QEMUFile *f, void *opaque, int version_id)
     return 0;
 }
 
+static void htab_cleanup(void *opaque)
+{
+    sPAPRMachineState *spapr = opaque;
+
+    close_htab_fd(spapr);
+}
+
 static SaveVMHandlers savevm_htab_handlers = {
     .save_live_setup = htab_save_setup,
     .save_live_iterate = htab_save_iterate,
     .save_live_complete_precopy = htab_save_complete,
+    .cleanup = htab_cleanup,
     .load_state = htab_load,
 };